LytchettNM

Hi newpersn sorry for not replying sooner been a little busy. Yes setup the 2 x DC's and get DNS & DHCP Working, the failover DHCP thing is easy to setup just have to remember if you need reservations you have to replicate between the servers for every change so setup one DHCP server first get all your bits setup correct then add the second DHCP, if I remember correctly if was a simple wizard. The key to getting all this working is getting the routing correct and ip helper / forwards setup on the device that is going to control your network, spend some time understanding this piece. Can you setup the new servers and a routing device without affecting your production network as you can get it right before moving everything over? When you have decided on the infrastructure setup of your server network you should then be able to take a look at the best place to put your FSMO roles, I opted to host my PDC with all FSMO on a VM which is hosted by hyper-v on a server which is not part of a cluster with two other SDC's one on each of our clusters to spread the load and support HA. I'll try to check back here and help when time allows if you need more help. Darren

All I can say is OMG a DC per VLAN crazy people setting that up for you. I set ours up last summer 3 server rooms 5 host servers across 2 clusters. It was a massive amount of work but we also choose to create a new domain as we moved from 2003 fictional level on 2008 servers to 2012r2. Still a little nervous about but everything into a failover cluster so we also added a server to host our primary domain controller within a VM which allows for us to backup the PDC, we then added 2 other dc's one within each cluster and it works very well. Our new core switching also does inter VLAN routing /ip helpers so we have VLAN's per building and cross site VLAN's for printing and wi-if etc. I would recommended failover DHCP across 2 server too with multiple scopes that way you can control all IP's from MMC.

Just some food for thought with hypervisor technology why pay for 4 hour support when you can cluster the servers and make key services highly available to cover hardware failure which is exactly what We have done 2 clusters across 2 server room with 2 server per room, all school services can run on just one cluster/server room. So hosts can be patched and if necessary we can add more host to increase capacity.

We have replaced all of our server's (25 HP) in the summer with 1u Lenovo rd550's which have been very reliable after some initial firmware issues, low power consumption, quiet and some very very good pricing for the spec.

The DUDE is awesome for free, Very configurable with a little time to work out how, 10 out of 10.

There are stacks of advice that can be offered but start at the basics, it looks like you have 3 edge switches and a switch you have setup as a core model numbers of the switches would be good and your ip range and subnet used would also be great. How many Ethernet ports do you have on the server/s backup drives and wifi controller again spec would help? Assuming that you have not changed the port configurations of the switches and they are all the same make linking the edge switches to the core switch with one cable then the ports should auto negotiate which is good, but if the switches are different makes then it might be better to fix speed and duplex on all the ports used to connect the switches together. I would also recommend enabling spanning tree if you have this option but only on ports facing workstations.

That's great news. Glad I could offer a little help. Regards Darren

The tester should with a little luck find the issue! If your structured cabling has been disturbed then I would say have a good look at that end, it's most likely a lose connection causing a high resistance with may be just one wire from that cable.

I would try re terminating the cabling at both the end of the suspect link making sure you cut a little off the end of each wire, I have seen many people just push down the termination tool over the top of the current termination and think that will sort it. Have you had any work done on the patch panel / cabinet that may have disturbed the cabling? But a good cable tester should be able to detect most of these type of fault and if you tester detects cable length comparing it to a socket next to the suspect one may help as that detects the resistance of the cable over its length so if wires inside the cable are broken you may get very different results.

A couple of questions to try and understand what you are hoping to end up with. Is the plan to set these new servers up as standard alone hosts and run VM's from each of them using the local storage or are you intending to cluster them to back each other up? Why no SAN? How many vServers are you hoping / need to run? What's the spec of the new servers i.e. core's, ram and storage space? What switching will they connect to? Do you understand LAG / LACP? Do you have core and edge switching in place and what speeds? Do you have a routing device within your network you control? Do you understand VLAN? Lots of questions but I'm hoping this will allow for a better overview of what you can achieve with the kit you have. And an idea of number of workstations and users could be useful. Darren

If you are having a duct installed between the two site's why not purchase a pre terminated fibre and pull it in yourself? Much cheaper we have run several on our campus.

Hi Phil Great news, I'm assuming you still have the ML310? If you are going to use Windows 2012 r2 then I would have the ML310 as a physical DC and a second DC hosted as a VM on ML350 DHCP and DNS on both DC's to split the load. Personally I would not use dynamic ram! I allow 4gb and 1 vCore for each of our VM's then add more if required. I agree with the 16GB for SIMS make sure you limit the max ram to about 12/14gb useable within SQL or SIMS will just grind to a halt and 4 core's for this one. Networking you have 8 ports to play with I the ML350 and as I don't know the storage box I'm assuming 4 port on that. 2 ports from the on-broad and 2 ports from the extra card into a lag for storage and connect directly to the NAS using say a 192.168.100.x/24 or less for network don't forget to check out jumbo frames to improve performance. Then on the client side of things use something like 10.42.x.x/24 network. For the servers I think I would use a pair of SSD's mirrored for the host's. Should be a nice little system. We run 6 servers, 1 hosting DC and network monitoring then the remaining 5 are arranged as 2 clusters in 2 server rooms to back each other up, each of the hosts has 2 X CPU's latest v3 Intel chip, 10 cores / 20 threads per CPU and 128GB per host. 1450 students and around 200 staff so just a little bigger.

You don't have to use a whole switch if you can use vlan or directly connect the server to the storage. In regards to you VM's you will have to assign vCPU's and ram to each VM as well as leave something for the Host, if you cut back to much the Performance of the system will just not be there. We run a min of 1 CPU & 4 Gb of ram per VM but some will require more resource like the SQL server you suggest maybe 2 vCPU's for that so a 4 core server will just not cut it as a host I'm sorry to say. If you are able to run the services you have listed on the one server already then I would stick with it until budget allows for more hardware. Regards Darren

I few things here, keep Storage traffic on its own network and interface if you can. You said you created LAG's have you used LACP? Had a quick check into the CPU and it looks like it's 4 cores? Don't raid the SSD's on the host server mirror them it will only be running the host OS. Storage drives, raid 5 or maybe 6 would be my recommendation. How many computers on the system. What IP range subnet are you using?